Re: parallelism

[Marcus G. Daniels]

>>>>> "RS" == Ralf Stephan <address@hidden> writes:

RS> What do you have in mind?  Some sort of log window? 

The first step is to get detailed logging information.  One problem
that exists now in Swarm is that methods can be complex and tracking
all the state change that occurs in a model is not possible just by
looking at the Actions that the activity library calls.  Here are some 
of the options we've come up with:
 
  1) Require that setter methods (instead of "=" assignments) be used
     to keep probing mechanisms in-sync.
       
     + easy to implement
     - requires programmer vigilance

  2) Rework Swarm (e.g. the Zone implementation) so that the MMU can
     can trap assignments and keep probing mechanisms in-sync.
  
     + fast
     + transparent
     - less portable

  3) Don't worry, be happy: make methods opqaue to probing, requiring
     modelers to do more fine-grained scheduling in order to expose
     state changes.

     + clear logging semantics
     - methods are opaque

Once there is a synchronized and efficient way to record `everything'
that happens in a model (including GUI button presses, etc.), then we
need to invent or integrate tools that provide metrics on the data.
We're just beginning to think about what to do in this area.  R has a
pretty good feature set to start with (cluster analysis, recursive
partitioning, etc).  They've got a COM interface and some Java
integration technology in the works, so I think R could be usable in
real-time adjacent to Swarm.



                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================